Rod Langman - 2:18am May 15, 1997 (#14 of 47)  

In response to Bill (#13), I'd like to add some additional items for consideration: in particular, the endogenous retroviruses that explain much of the old Mls-locus effects of T cells (where deletion is simple and clear) and the case of adults that have been lethally irradiated and reconstituted with antigenically different bone marrow, and again, solid tolerance is achieved. It seems from the irradiation type of experiment that regenerating an immune system in the absence of functional T and B cells is a way of recapitulating embryology. The temptation to look for self markers has to be resisted because they would not easily explain the immunity you get when the near-self serum globulins are mixed with LPS and bacterial products in adjuvant (the markers would end up being as specific as antibodies and TCRs). 

The various transgenics that have been constructed express all kinds of strange "self," ranging from viral antigens to allo-MHC to hen egg lysozyme, etc., and provide examples of almost every imaginable outcome. Some transgenes are expressed late and treated as nonself, and others are expressed at such low levels that the immune system ignores them. The latter cases of low-level expression include examples where it can be shown that the transgenes are adequate targets for immune elimination without being inductive. In this case, there seems to be a higher sensitivity to immune elimination in the heat of a response, whereas the sensitivity for induction of tolerance seems somewhat higher. All this does is remind us that immune responses are concentration/density dependent, with some kind of threshold for induction/tolerance and another for effector function. There is also a reminder that birth is not necessarily a significant marker (for example, sheep in utero make decent responses to many antigens, whereas mice do not). Thus, whatever primary mechanisms are postulated, none are going to avoid the law of mass action, and it is going to be difficult to expect a single explanation for all events - the surprise of knockout mice being less affected than predicted many times is a good lesson.  

My question is, What determines whether an antigen is going to be processed and cause the expression of stimulants, such as inflammation and "danger"? Bill, you see many substitutes for the so-called signal[2], and this pretty much rules out any signal[2], and you remain unconvinced by Ephraim's necrosis/apoptosis argument, but what do you imagine is in control of induction aside from self markers? It seems that we have many ways to enhance an immune response in adults but few ways to dampen them specifically. 

Rod Langman - 2:27am May 15, 1997 (#15 of 47)  

Antonio, although I think I have an answer for your questions (#10), it is not obvious that my having an alternate answer would convince you to change your model. In the end we would quibble over whether the answers were good or bad, and this would in turn rest on how we interpret the other's model. Thus, I'd like to cut to the chase and get our different positions made crystal clear so that everyone knows where our respective models agree and disagree.  

Rod Langman - 9:08am May 15, 1997 (#16 of 47)  

It seems that exhaustion is setting in, as responses are shorter and slower. So, rather than wait, I decided to put on an alarmist's hat and try to argue the points as they might, putting aside questions of exactly who said what and where for the sake of clarity and brevity.  

The alarmist speaks: 

One of the main arguments against the original two-signal model is that it explained how tolerance and immunity could be maintained but failed to explain how to establish a self-nonself discrimination. This was apparent to all from the beginning, just as it was apparent that the Lederberg model, as stated by Lederberg, could not deal with a regenerating immune system. 

Then came the updated AAR model, which picked the best bits of the old ones, lifting an antigen-independent differentiation step from Lederberg and pasting it on to the iTh lineage to generate eTh that were then in reversible equilibrium with some steady-state levels of iTh and eTh. From the Bretscher-Cohn two-signal model came almost unchanged how to maintain the embryonic determination of self and nonself as outlined here. The criticism against this is along the lines that not all iT and iB cells have an obligatory requirement for eTh, and so the model is inadequate. 

A general criticism of all models of self-nonself discrimination is that antibodies to self components can be found in normal healthy individuals, and this is a denial of an independent self-nonself discrimination mechanism. 

What does the AAR model have to say?  

From a purely evolutionary point of view, there was a time in the past when antigen simply induced a germ-line-selected effector reaction; one of the ways of ensuring that it was directed at a pathogen was to make the threshold of induction high so that at very high antigen doses, when inflammatory reactions were maximal and essential, there was a pathway of "help"-independent induction. I see no reason for evolution to have eliminated such a pathway, even in the immune system of today. Rather, I would argue that immune regulation, like the antibody pathway of complement fixation, simply lowered the thresholds for signaling, and what is referred to as signal[2] (the active principle of an eTh) is probably not much more than a concentrated slug of whatever the original signal was. Thus, to find pathways for overriding signal[2] in special circumstances is hardly surprising. One of the special circumstances is infection by viruses that infect cells of the immune system, along with bacterial adjuvant-like materials. However, the simple, now old-fashioned "hapten-carrier" effect showed that antigen-specific T-cell help was required for a B-cell response; similar demonstrations have been made by Jim Forman and later by Polly Matzinger for cytotoxic responses to the Qa1 alloantigen using help to H-Y male antigen on the same cell. The net result is that whichever position one takes, there is evidence that eTh are not required, and eventually both cases have to be accommodated. With regard to antibodies to self components, we have seen that making a genetic definition of self leads to absurdities such as this type of argument. Only the immune system's definition is meaningful to the immune system.  

Why the alarmists rarely attack the dominant-suppressive self position is unclear; either they disregard this view, or they see no contradictions.  

The alarmist speaks again: 

Well, if you have to resort to hand waving like this, why not wave hands over alarm signals, because they explain the phenomena too. So, let me construct a minimal model.  

T cells are the regulators of immune behavior, and if their activity can be accounted for, then all the rest of the cells will fall in behind. T cells begin life in the thymus, whether the animal is embryonic or adult, and here they undergo "negative selection," meaning that antigen interacting with the T cell (the antigen is a processed peptide cradled in an MHC molecule) will cause elimination of that T cell. This process of eliminating T cells can be thought of as a Lederberg-style event in the sense that these thymic T cells can only be inactivated by antigen (we will ignore positive selection in the thymus). Thus T cells leaving the thymus have been purged of many (most?) self-reactive cells, at least in the case of self antigens present in thymus. When these T cells exit the thymus and enter the periphery, some antiself reactivity remains because not all self antigens can be present in thymus, even if purging of all cells reactive to thymic self is perfect. These would be iT cells that are capable of being either induced to immunity or tolerized (thymic T cells probably cannot be induced to immunity). The problem is now how to generate the first eTh and kick the immune response into high gear.  

Rather than use antigen-independent steps in antigen-specific cells, alarmists look to the infectious, invasive, or disintegrating forces as a source of stimulants that in some models (danger and integrity) come from the damaged self tissues, and in Janeway's model the stimulants come from the pathogen itself. In other words, either the pathogen or its pathogenic effect is the source of triggers that produce eTh in an antigen-dependent step. Despite the many flavors and directions of individual signals, they can be lumped for the sake of analysis of principle, not detail. This class of models would allow antiself T cells that sneaked out of the thymus to be inactivated upon encountering antigen in the absence of an alarming event. But, in the present of alarms, all antigen-reactive cells, whether antiself or antinonself, would become inducible to immunity. No matter that some antiself in the vicinity is induced, because this is not going to kill more cells than the pathogen itself, and when the pathogen has been cleared, no alarms will be sounding, and any remaining antiself will continue to be inactivated. 

Because the killing of virally infected cells is itself a potentially alarming event, and any bystander antiself that killed cells would become equally alarming, the immune effector mechanisms cannot themselves be alarming - hence the distinction between immune apoptosis (a quiet natural death) and necrosis (a violent unnatural death). Thus, when the pathogenic inducer/producer of alarm signals is eliminated, the immune system returns to rest, despite any small local outburst of autoimmunity. 

What does the AAR model have to say?  

First, I'd want to be among those who argue strongly that inflammatory responses to infection are essential and play a vital, even pivotal role in many (most?) cases. But is this good enough to make an effective self-nonself discrimination to the point of explaining the behavior of the normal immune system? 

The fundamental principle of alarm, whether induced or produced, requires a germ-line-selected system to recognize these signals in all animals, syngeneic or allogeneic. In particular, these signals do not assort like antigens recognized by T and B cells. Thus, at this level we are dealing with a self-nonself-marker class of model, and these are generally considered inadequate as a matter of principle. However, the alarm system is really only a backup to deal with a few straggler antiselfs that escaped the real site where the self-nonself discrimination is made, namely the thymus. There are several ways to manage the thymic problem. One can use a tolerance-only stage, or one can find ways to exclude pathogens and various sources of alarm to impose a de facto tolerance-only environment (i.e., free from alarm). I would add that under the AAR model the thymus becomes a mostly special place because it is where iTh are born, and it is not until well after iTh leave the thymus that they can become antigen-independent eTh, making the thymus a site uniquely deficient in eTh. Thus, by two different pathways we would reach similar phenotypes.  

But just how good is alarm at dealing with the "facts"? It seems that for every fact fished out in support of one model, there is an equal and opposite antifact that can be found. I've mentioned before, and no one has answered (my alarmist hat fails me here), how antigenically different grafts are rejected when antigenically identical grafts take. How does one conjure alarm signals here? Turing to the tumor example raised in the "danger" case, in message 8 Ephraim drew attention to epithelial tumors being tolerogenic because they lacked alarming dendritic cells. My reading of the literature suggests that fibroblasts are not dentritic cells either, and when tumors are removed, the same tumor cannot be reimplanted. Transferring tumors from one animal to the next requires quite large cell inocula, and one has to wonder why it took so many cells to establish a tumor. My guess is that tumors are a lot like other pathogens: They are in a race with the immune response, and whoever gets there first wins. A large tumor inoculum gives the tumor a head start. With respect to primary tumorigenesis, and how tumors avoid immune elimination, I'd rather deal with that on another occasion, as it is not strictly germane to the subject at hand (tumors generally kill old and postreproductive individuals).  

Finally, if alarms are such good indicators of the thingies to be eliminated, why bother with all the complicated immune specificities when there is a perfectly good alarm-recognition system that can kill all in sight until the alarms go off? After all, we take for granted a huge evolutionary investment in cells, an entire parallel plumbing system with the lymphatics, and exquisitely organized organs such as thymus, spleen, and lymph nodes, not to mention the huge genetic cost of immunoglobulin and TCR gene banks. Why such a complicated system to do an alarmingly simple job? There is good reason to believe that we are under a constant steady-state pathogenic load that is not sporadic like winter coughs and colds, mixed with occasional bouts of food poisoning (we call it Montezuma's Revenge here). Rather, it is probably a constant daily load that is eliminated promptly, unawares to us. But this steady-state pathogenic load is also a steady-state source of alarm and carries a nontrivial potential for chronic persistent autoimmunity.  

I will offer one further posting to deal with the question of antigens near and far from self, as Bill is constantly reminding me must be dealt with by all models. 

Rod Langman - 6:42pm May 15, 1997 (#17 of 47)  

For completeness, some comments are in order concerning the question of cross-reacting antigens, ones that contain both self and nonself epitopes. One nice example of these are the serum globulins that Bill described as excellent tolerogens. The principle is the same as discovered by Landsteiner more than 50 years ago. He saw a gradient of cross-reactivity between the same class of proteins taken from species near and far from the animal being immunized. Interestingly, he was stumped by the finding of almost all-or-none reactions with cellular antigens (e.g., the famous ABO blood groups). We understand the cellular antigen observation in terms of the carbohydrate antigens, which are all-or-none sugar additions, compared with amino acid substitutions making individually small changes most of the time. 

For simplicity, consider an antigen made up of just two kinds of epitope, S and F, symbolized as SF. These can be serum globulins that are not alarming pathogens. We know that the closer the donor species, the closer the ratio of S to F epitopes approaches 1:0, and the further away the donor, the closer the ratio tends to 0:1. We also know that antigens close to self are better tolerogens than those distant from self, and conversely with respect to immunity. This kind of observation is not easily explained by the alarmists unless they admit a graded expression of a self marker that can be quantitatively counted so as to measure phylogenetic distance. The "networkists" with the dominant-suppression principle at work would probably say that the number of epitopes connecting an antigen to the network defines the strength of the pull one way or another. Under the AAR model, there is a similar idea at work based on the anti-F i-state cell receiving a number of eTh signals dependent on the amount of F displayed. In other words, there is a ratio of signal[1] to signal[2] that approximates to the ratio of F to S. The iTh makes its decision based on the ratio of the signals; statistically averaged in a large population of cells, the probability of any one cell being sent to tolerance or immunity depends on the ratio of the signals, and the total number of cells engaged depends on the total number of epitopes available. As an aside, there is no difference in principle for T cells that only recognize processed peptide displayed on MHC molecules. 

The tricky part comes when tolerance is broken by a cross-reacting antigen. I'm pretty sure Bill was the first to show unequivocally that tolerance he induced to one protein could be broken by immunizing with a cross-reactive protein, including a strongly haptenated derivative of the tolerogen. The first antibodies to appear when tolerance was broken were those to the common epitopes, and later antibodies unique to the tolerogen could be found. Again, the alarmist view is inadequate because it would predict, as best I can determine, the simultaneous appearance of antibodies to the shared and the tolerogen-specific epitopes. The dominant-suppressionist view is not clear because I'm not at all sure how to keep the immune and suppressive networks apart in the absence of antigen, and I'm even less sure how to have them interact in the presence of antigen. It would be nice if we (or I) could get a brief outline of how this works in principle - no fancy details needed. 

Under the AAR model, the breaking of tolerance ultimately depends on newly arising i-state cells, say iB, specific for the tolerogen (including those specific for epitopes shared with the cross-reacting antigen). Tolerance is the continued lack of eTh cells that were eliminated in the tolerizing event by means that might be discussed another time. Newly arising iT and iB specific for the tolerogen interact with persistent tolerogen and, in the absence of eTh, are sent packing down the pathway to the gallows and certain death. It would be reasonable, under these conditions, to consider the tolerogen to be the equivalent of an S antigen. Along comes the cross-reacting antigen SF, which can react with eTh and iT and iB that are anti-F. There is an expected response to the F epitopes unique to the cross-reacting antigen. The newly arising iB specific for S in the SF complex can be helped when they present F obtained from an SF complex, but they also obtain plain S from the persistent tolerogen. Thus, there is a competition for the iB cell to bind S and SF, and because the choice of tolerance or immunity in this iB cell is dependent on the ratio of signal[1] (via S) to signal[2] (via F), it will be possible to break tolerance with sufficient signals from eTh anti-F. Indeed, Bill has shown that adding more tolerogen blocks the breaking of tolerance by the SF complex. As Bill also has pointed out, T cells have a lower threshold of detection of antigen than B cells, and so it will be the B cells that first escape tolerance as the level of tolerogen diminishes and T cells last. Thus, eventually iTh specific for S will be able to become eTh when helped by the eTh that are anti-F. Only when these anti-S eTh are available does the system make a response to epitopes unique to S (the tolerogen). This is because only the tolerogen can induce a response to the epitopes unique to the tolerogen.  

As a general rule, there are only two ways to break tolerance to a true self component under the AAR model. One is via the cross-reacting antigen, as illustrated above; the other has more to do with protecton theory and depends on leakiness in the allelic-exclusion (one receptor specificity per cell) rule of clonal selection. Some cells will have two receptors, one anti-S and another anti-F, and if the cell can be induced via anti-F, the anti-S will be entrained and expressed so long as the F agent persists.  

The induction of tolerance in adults is a matter of finding ways to eliminate eTh without letting them express their effector function. Typically, it is impossible to render tolerant an animal that has been immunized, even quite a long time ago, unless one resorts to lethal irradiation, and even then it is hard. The best guess is that the cycling between eTh and iTh in the absence of antigen creates an opportunity to gradually sneak the iTh off into the woods and kill them without engaging the eTh, and as eTh revert to iTh, they too can be carted off. The introduction of massive amounts of antigen that are poorly processed keeps a steady supply of low levels of processed antigen spread uniformly throughout the system. Because the eTh only interacts with iTh via a processed antigen intermediate, the probability of an eTh being present at the same time an iTh reacts with antigen is low, and slowly the iTh will be driven to tolerance. The iB cells react with antigen without any processing and receive the full dose of soluble antigen, leaving the ratio of signal[1] to signal[2] heavily in favor of signal[1] and tolerance.